ECG Arrhythmia Detection with Wavelet Features and Machine Learning

Every heartbeat of a 30-minute Holter ECG is classified into the AAMI classes N (normal / bundle branch block), S (supraventricular ectopic), V (ventricular ectopic) and F (fusion). The project uses the real MIT-BIH Arrhythmia Database (PhysioNet): a 0.5-40 Hz Butterworth band-pass, an own Pan-Tompkins R-peak detector (sensitivity 99.69 %, positive predictivity 99.13 % on all 48 records), a 253-sample beat window, db4 level-4 wavelet coefficients plus RR-interval features, and Random Forest / SVM classifiers compared with 1-D CNNs. It is evaluated honestly with the de Chazal inter-patient protocol (train on 22 patients, test on 22 others) next to the misleading intra-patient random split, tested on an external database and under real electrode-motion, muscle and baseline-wander noise. WFDB readers and GNU Octave reproduce the signal chain; a complete MATLAB version, an Arduino + AD8232 front end and a web application with five databases are included.

Category: Electronics and Communication Engineering (Biomedical Signal Processing, ECG, Pan-Tompkins QRS Detection, Discrete Wavelet Transform, Machine Learning, 1-D CNN, WFDB / PhysioNet, GNU Octave, MATLAB, Python)

Electronics & Communication Engineering Project E06 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.